This project analyzes disordered control of posture and movement during reaching in patients with stroke resulting from middle cerebral artery occlusion. Prior studies have shown that control of posture and movement may both be impaired, and contribute importantly to disability. Our recent work has shown that normal reaching requires control of both posture and movement, that these two functions are specified by different neural mechanisms, and that accuracy is dependent on adaptation (learning) and coordination between the two. Based on our preliminary observations, we hypothesize that stroke degrades the ability to recruit and relax the balanced muscle co-contractions needed both to overcome limb postural bias associated with hypertonia and to stabilize the hand against unsteady loads across the workspace. We further hypothesize that major impairment in reaching post-stroke arises from improper scaling and timing of muscle coactivation, thus limiting the independence of arm trajectory and position control. Using a planar robot arm and novel EMG biofeedback methods, Aim 1 proposes to characterize the stroke-related changes in: 1) the ability to maintain postural stability throughout the workspace via graded coactivation of antagonist muscles;2) the latency and developmental time course of different levels of coactivation, and 3) the use of proprioceptive feedback to counteract reflex abnormalities during regulation of limb position in the presence of mechanical perturbations. Aim 2 examines stroke-related changes in the integration of posture and movement control. Using the planar robot we will: 1) explore transfer of learning between posture and movement tasks, assessing whether coupling between the two controllers increases post-stroke, and 2) determine whether trajectory planning adapts post-stroke to account for the biomechanical effects of posture regulation. We expect our results will ultimately lead to new approaches for rehabilitating arm function post-stroke. We anticipate that improvement in arm function will be promoted by applying robotic and biofeedback methods first to train coactivation, posture and movement control separately, then in combination.